666ghj/MiroFish · error · ValueError

脚本不存在: {script_path}

Error message

脚本不存在: {script_path}

What it means

Raised by SimulationRunner.start_simulation when the OASIS simulation script path passed in the config does not exist on disk. Before raising, the runner marks the run FAILED, clears twitter/reddit running flags, appends any Zep graph-write cleanup failure to the message, persists run state, and syncs status. The ValueError surfaces the composed state.error string to the caller.

Source

Thrown at backend/app/services/simulation_runner.py:500

                try:
                    ZepGraphMemoryManager.stop_updater(simulation_id)
                    cls._graph_memory_enabled.pop(simulation_id, None)
                except Exception as error:
                    cleanup_error = error
            state.runner_status = RunnerStatus.FAILED
            state.twitter_running = False
            state.reddit_running = False
            state.error = f"脚本不存在: {script_path}"
            if cleanup_error is not None:
                state.error += f"; Zep图谱写入清理失败: {cleanup_error}"
            with cls._finalization_lock(simulation_id):
                cls._save_run_state(state)
                cls._sync_simulation_status(
                    simulation_id,
                    RunnerStatus.FAILED,
                    state.error,
                )
            raise ValueError(state.error)
        
        # 创建动作队列
        action_queue = Queue()
        cls._action_queues[simulation_id] = action_queue

        process = None
        main_log_file = None

        # 启动模拟进程
        try:
            # 构建运行命令，使用完整路径
            # 新的日志结构：
            #   twitter/actions.jsonl - Twitter 动作日志
            #   reddit/actions.jsonl  - Reddit 动作日志
            #   simulation.log        - 主进程日志
            
            cmd = [
                sys.executable,  # Python解释器

View on GitHub (pinned to b5b53acc57)

Solutions

1. Verify the exact path in the error message exists on the machine running the backend (ls the path from the message).
2. Make script_path absolute (or resolve it against a project-configured base directory) before calling start_simulation.
3. Re-generate the simulation config so script_path matches the current script location shipped with the deployment.
4. If the '; Zep图谱写入清理失败' suffix appears, also fix the Zep connection/key issue reported there — the run failed for both reasons.

Example fix

// before
config.script_path = "scripts/run_twitter_simulation.py"; // resolved from unknown CWD

// after
import app_config
config.script_path = str(
    (Path(app_config.PROJECT_ROOT) / "scripts" / "run_twitter_simulation.py").resolve()
);
if not Path(config.script_path).exists():
    raise FileNotFoundError(config.script_path);

from pathlib import Path

def can_start(config) -> tuple[bool, str]:
    p = Path(config.script_path)
    if not p.is_absolute():
        return False, "script_path must be absolute"
    if not p.is_file():
        return False, f"script missing: {p}"
    return True, ""

def is_startable_config(config) -> bool:
    p = Path(getattr(config, "script_path", ""))
    return p.is_absolute() and p.is_file()

try:
    SimulationRunner.start_simulation(cfg)
except ValueError as e:
    if str(e).startswith("脚本不存在"):
        # surface as config error, not server error
        raise HTTPException(400, detail=str(e))
    raise

When it happens

Trigger: Calling start_simulation (or the API endpoint that wraps it) with a config whose script_path points to a missing/unmounted file; the runner builds the full path from config and the os.path/existence precheck fails. The message may also carry '; Zep图谱写入清理失败: ...' when the cleanup of a previously started graph updater also errored.

Common situations: Relative script paths resolved against the wrong CWD (backend started from repo root vs backend/), renamed or moved simulation scripts, deployment where the scripts directory is not shipped or mounted, or a stale config JSON referencing an old script name.

Related errors

• 等待命令响应超时 ({timeout}秒)
• 模拟不存在: {simulation_id}
• 没有找到符合条件的实体，请检查图谱是否正确构建
• 不支持的平台: {platform}
• 模拟配置不存在，请先调用 /prepare 接口

AI-assisted analysis of 666ghj/MiroFish@b5b53acc57 (2026-08-14). Data as JSON: /api/errors/964993d5aff7ba90. Report an issue: GitHub.